|
Posted by MrMark501 on 12-15-2004 6:38 AM
|
|
We recently had a problem on our board designs, specifically in the schematic, where a symbol was changed at rev 2 of the board and a replace cache was done in order to update the symbol. During the rev C spin of the board the cache again had to be updated due to the libraries being relocated to a different drive on the server. The symbol was not changed locally because there was no internal revision on the symbol in the design cache (the presence of an underscore followed by a number) This caused the new revision of the board design to be connected up per the old symbol, not the new, corrected one. Consequently, the boards arrived and when powered up, severely damaged one of the chips. Has anyone had this appen to them or know of any likely scenerios where this could happen?
Thanks
|
|
|
|
Posted by rmann on 12-15-2004 7:48 AM
|
|
Sorry that I can't shed any light on your question. But I need to point out that the word is "part" not "symbol". In Capture the term "symbol" is reserved for:
-Power/ground
-Offpage Connector
-Hierarchical Port
-Titleblock
-Richard Mann
|
|
|
|
Posted by Ron_O on 12-15-2004 10:08 AM
|
|
I don't really see how this could happen unless the libraries that got moved were the old (unedited) libs. It sounds like you edited the symbol in the library -correct? Then did a Replace Cache to update the symbol in the Design cache. Assuming you didn't change the name of the edited symbol, the Design Cache should have the same name as the lib symbol (and have it's edited features). Have you checked that the relocated lib has the symbol(s) updated? Also, do the part properties list the correct Source Library reference?
You didn't mention the version of Capture you're using. Perhaps it's a new 'feature' if it's v10 :), that retains the old symbols instead of wiping them from the design database. Very strange (and disturbing) behavior if that's the case.
A situation where it's good policy to print out all ECO changes before layout update proceeds and have the engineer review any that involve footprint or connection change.
Ron O.
|
|
|
|
Posted by MrMark501 on 12-16-2004 6:00 AM
|
|
rmann,
I was using the term "symbol" because around here when we talk about symbols we're talking schematic and "footprint" when we talk about PCB parts. the term "part" tends to confuse people so it's just habit we've formed when talk components.
Ron_O,
The scenerio you brough up is one we considered, but the IT person copied the library directly from one drive to the other while noone was in it. We are using 9.2.3 on the capture side for most of our engineers because I don't think we've upgraded the licenses, but we have determined that 9.2.3 will let more than one person in the same symbol in the library at the same time. My theory is that 2 people had it open, 1 vieweing and 1 editing. The editing person saved first and update the cache. The second person inadvertantly saved after the cache was updated, thus negating the change in the library, not the board. Version 10 has file locking. The second person would have gotten a read-only message if we were using version 10. I think the "feature" :0) you speak of was actually in 9.2.3. There is also the possibility that the person editing the symbol made the change and updated the cache. THe the computer crashes before the library can be saved. When Orcad is called up again, the recovered library doesn't have the edited part, but the old part.
Just some thoughts. Still interested in any theories anyone might have. I just can't explain it too many other ways.
Later,
Mark
|
|
|
|
Posted by Ron_O on 12-16-2004 12:15 PM
|
|
Mark,
I think your 1st theory of 2(or more) users with the symbol open for editing is a likely senario. I've heard of other users having revision control problems in 9.2 -mostly in design level though.
Another possibilty here is that the symbol was edited at design level. This does add an underscore suffix to the symbol name in the design cache but the Replace Cache name could have been altered to remove the suffix. Then when the update was done after lib relocate the original (unedited)symbol would overwrite the Design cache one. In this case it comes down to: was the symbol edited at the lib level or design level?
If you've got several engineers working (editing) the libs you could have future problems as well. You might want to consider some basic revision control like making the libs read-only access to all but one controller who would gather local symbol updates and update the master lib at specific times. It's extra work for someone in the chain but better than hardware failures and more board spins.
Ron O.
|
|
|
|
Posted by MrMark501 on 12-17-2004 6:08 AM
|
|
There's no underscore so we are surmising that it had to come out of a library. Thanks for your theory about removing the suffix. We're pretty sure this one didn't happen at the design lever. All the engineers have been burned on that more than once.
We've since made it a policy that the engineer can use new symbols in the library, but a design group person has to actualy enter it in the library. There's just to many for one person to handle and still get designs out the door.
Thanks,
Mark
|
|